Surgical Samples and Demographics of Participants

The Cancer Institute of Tianjin Medical University and the Hospital Ethics Committee (bc20240074) approved the use of all samples and participants’ information. All participants provided written consent for the use of their samples and disease information for future investigations according to the ethics committee and in accordance with the recognized ethical guidelines of Helsinki.

Sorting of Human and Murine CAFs

The Ethics Committee approved the acquisition of all human PDAC samples from the Pancreatic Cancer Department of Tianjin Medical University Cancer Institute & Hospital. These samples were collected with the informed consent of the donors. The clinical information of participants whose tumors were used for isolating CAFs is presented in Table S1. Human pancreatic CAFs were isolated from fresh PDAC surgical samples using a culture outgrowth approach. Specifically, fresh human PDAC surgical samples were sliced into blocks of 1–3 mm using a sharp blade. These blocks were subsequently seeded in 6-cm culture dishes and cultivated with DMEM medium supplemented with 10% fetal bovine serum. After 7–15 days, when the cell confluence reached 90%, CAFs were subjected to trypsinization, replated into another culture plate, and further cultured with a complete DMEM medium. The isolated cells were identified using immunofluorescence and FCM, showing positivity for the following mesenchymal-specific markers: Desmin, Collagen I, and α-SMA, and negativity for other cell lineage markers: CD326 for epithelial cells, CD31 for endothelial cells, and CD45 for immune cells. The human PDAC cell line SW1990 was kindly provided by Prof. Keping Xie (MD Anderson Cancer Center, Houston, TX) and was maintained in RPMI-1640 with 10% FBS.

Screening of Primary Resting Fibroblasts from Human and Murine Sources

(1) The pancreas of mice was removed and cut into small pieces and rinsed with PBS; then the floating fat was removed. (2) Digestion: 0.02% collagenase IV, 0.02% pronase, and 0.05 DNAse were added to the enzyme dilution to a final volume of 10 ml. 3). The digestive enzymes and tissue samples were mixed evenly and incubated in a 37 °C water bath for 12–15 min. 4). Then 10 ml of complete medium was introduced to the aforementioned volume to counteract the digestion and subsequently underwent filtration using a 100-µm filter. 5) The resulting mixture was centrifuged at 450 g for 7 min. Subsequently, the supernatant was removed, and 5 ml of 15% optiprep (1.25 ml optiprep + 3.75 ml tris base-NaCl) was added. The tube wall was gently brushed with 5 ml of 11.5% optiprep (0.96 ml optiprep + 4.04 ml tris base-NaCl), and 5 ml of GBSS was gently brushed along the tube wall. 6) Then it was centrifuged at 1400 g for 17 min and the white film layer between GBSS and 11.5% optiprep was collected. 7) The collected white film layer was re-suspended and washed with GBSS, centrifuged at 450 g for 8 min, and cultured in a complete DMEM medium.

ScRNA-Sequencing Dta Processing

The 10X Genomics-based scRNA data for 35 PAAD samples were downloaded from the Genome Sequence Archive (CRA001160). A total of 57,004 cells were annotated and distinct cell clusters were identified using representative markers. The default parameters of the “Seurat” package were utilized, and the harmony algorithm was employed to de-batch the data. Subsequently, the Uniform Manifold Approximation and Projection (UMAP) technique was applied to reduce the dimensionality and visualize the cell subpopulations. The differentially expressed genes (DEGs) were compared among cell clusters with the Seurat ‘FindAllMarkers’ function. The fibroblasts were divided into seven subpopulations according to highly expressed genes. The “CellChat” package was used to infer and analyze cellular communication from the PAAD scRNA-seq data. The pseudo-time evolutionary trajectory was inferred from the Slingshot method.

Development of RNA-Sequencing Library

The strand-specific libraries were developed by using the TruSeq Stranded mRNA Sample Prep Kit (Illumina), according to the manufacturer’s instructions. Poly-adenylated RNA from intact total RNA was refined using oligo-dT beads. The extracted complementary DNA fragment was ligated to Illumina paired-end sequencing adapters, 3′ ends adenylated and amplified by PCR. The libraries were sequenced with 65-base pair (bp) single-end reads on a HiSeq 2500 System in high output mode using V4 chemistry (Illumina). Gene expression levels were quantified by Salmon (see URLs) using default parameters for both applications after raw reads were aligned to GRCh38 using a STAR RNA-seq aligner.

Genomic Mutation Analysis

The somatic mutations in the Mutation Annotation Format (MAF) file for all the PAAD patients were obtained from the TCGA cohort. Maftools is agnostic of bigger alignment files and simply requires somatic variations in MAF [46]. It provides many analysis and visualization modules, including driver gene identification, signature, pathway, enrichment, and association analyses, that are frequently used in cancer genomic investigations. This study employed the R Bioconductor package “Maftools” package to analyze and visualize PAAD genomic mutation data. The “Forestplot” package was also used to compare genes with significant mutation frequency differences between the two groups.

Analysis of Immune Cell Infiltration

The GSVA scoring system was employed to calculate the infiltration scores of highly expressed genes in MME+CAF for each sample in the bulk transcriptome to determine the MME+CAF infiltration scores. Based on the median of the infiltration scores, they were categorized into high and low groups. A diverse array of algorithms, such as TIMER, CIBERSORT, CIBERSORT-ABS, EPIC, MCPCOUNTER, XCELL, and QUANTISEQ, were implemented to quantify immunoreactivity to compare the immune microenvironment of high and low MME+CAF infiltration groups. The degree of correlation between MME+CAF and immune cells was calculated using Pearson correlation analysis. The TIDE algorithm was also employed to evaluate the immune exclusion score, immune dysregulation score, and total TIDE score in different groups.

Prediction of Drug Sensitivity of MME+ Fibroblast

The cell line expression data were obtained from the Genomics of Drug Sensitivity in Cancer (GDSC) database (https://www.cancerrxgene.org/). The IC50 value matrix was also included for drug sensitivity analysis. The “oncopredict” R package was employed to generate predictive drug sensitivity data (IC50) for pancreatic cancer samples. The Connectivity map analysis (CMap) online tool (http://clue.io) has been verified to be useful in silico drug screening tools to target disorders. The genes positively correlated with MME+fibroblasts were subjected, and the query result was a list of drugs with a “connectivity score” ranging from + 1 (positive connectivity) to -1 (negative connectivity).

Cell Culture and Transfection

Primary CAFs were isolated from invasive pancreatic ductal adenocarcinoma samples obtained from surgery or KPC genetic engineering mice. The KPC mice were primarily 10 weeks old and were initially identified through PCR for KRAS-TP53 mutations, and subsequently through in vivo ultrasound for tumorigenesis and tumor burden. Briefly, tissues were digested by collagenase type I, collagenase type III, and hyaluronidase (1.5 mg/ml) at 37°C with agitation for 2–3 h in DMEM with 10% FBS. Subsequently, the dissociated tissues were incubated at room temperature for 5 minutes without shaking to isolate primary fibroblasts. Then the stromal cell-enriched supernatant was removed and transferred to a new tube. Human fibroblasts were then cultured in DMEM with 10% FBS and the purity of fibroblasts was validated by flow cytometry analysis. CAFs from passages 2 to 10 were used for subsequent experiments [47]. To introduce genetic material into CAFs, a total of 1 × 105 CAFs were placed in each well of 6-well plates. Lentiviral particles with a multiplicity of infection (MOI) of 100 were added to the CAFs, along with 5 µg/mL of Polybrene. The CAFs were then incubated with these substances for 12 h. To transduce mouse fibroblasts, 5 × 105 cells per well were treated with mouse MME-overexpression lentiviral particles (MOI of 20) and 5 µg/mL Polybrene for 12 h in 6-well plates. The human pancreatic cancer cell line (SW1990) and the murine pancreatic cancer cell line (PANC02) were cultured on RPMI1640 and DMEM media supplemented with 10% Fetal Bovine Serum (FBS), 100 U of penicillin, and 100 µg/ml streptomycin at 37℃ in a humidified atmosphere of 95% air and 5% CO2.

Western Blot

Cells were lysed using RIPA lysis mixed with phosphatase inhibitors and protease inhibitors. Protein concentrations were measured by BCA protein assay. The same amount of protein was separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Related antibodies were diluted at 1:1000 and incubated overnight at 4 °C with membranes containing blot proteins. Horseradish peroxidase-conjugated goat anti-rabbit antibody and goat anti-mouse antibody were diluted at a ratio of 1:5000 as secondary antibodies. The blots were detected with a Chemi-Scope exposure machine. We have added the concentration and source of the protein in Table S2.

qPCR

Total RNA was extracted from pancreatic cancer cell lines using TRIzol reagent. Then, the mRNA was used for first-strand cDNA synthesis with the Reverse Transcription PCR System (Bimake) according to the manufacturer’s instructions. RT-PCR was performed to measure the mRNA levels of the target genes. Each RT-PCR experiment was repeated independently at least three times. β-Actin was also used as a loading control. The primers information was supplemented in the Table S3.

Flow Cytometry

Flow cytometry is mainly based on mouse-derived sorted CAF and KPC cell lines mixed and injected subcutaneously into immunocompetent mice to form tumors. The tissues contain KPC cells, CAFs, and many immune cells. For mouse tumor tissues, samples were prepared into a single-cell suspension. The harvested cells were divided into separate tubes for each antibody staining. After adding appropriate concentrations of fluorochrome-conjugated antibodies, the mixture was incubated for 30 min away from light. The above samples were detected using a Beckman flow cytometer and the obtained data were analyzed in software Flow Jo-10.0.

Multiplex Fluorescent IHC

The development of multiplex fluorescent IHC is mainly carried out after subcutaneous tumor formation by mixed injection of mouse-derived CAF and mouse-derived KPC cell lines for sectioning. In brief, 5 μm of PDAC slides were deparaffinized and rehydrated through a graded series of ethanol solutions (100% 1 × 10 min; 100% 2 × 10 min; 95%×10 min; and rinsed in 70%) before antigen retrieval in heated Citric Acid Buffer (pH 6.0) in microwave treatment for 15 min (EZ Retriever microwave). Each slide was put through the process of staining including a protein block with blocking buffer followed by primary antibody targeting Ki67 and corresponding secondary HRP-conjugated polymer. Each HRP-conjugated polymer mediated the covalent binding of a different fluorophore for signal amplification. Ki67 was labeled by Opal 690(676–694 nm). This reaction was followed by additional antigen retrieval in a heated Citric Acid Buffer (pH 6.0) for 15 min to remove bound antibodies. DAPI was used to identify nuclei, and slides were mounted with a fluorescence mounting medium. All analyses were conducted with the same type of control. Images were captured by the Zeiss fluorescence microscope (400×) and analyzed in ImageJ.

TUNEL Staining Assay

The tumor tissues were fixed in formalin, embedded in paraffin, and cut into sections of 5 μm. TUNEL assay was used to detect apoptosis in the tumor tissues according to the manufacturer’s protocol. Tissue sections were analyzed to detect the localized green fluorescence of apoptotic cells, and DAPI was used to visualize the cell nuclei. Finally, images were captured by the Zeiss fluorescence microscope (400×) and analyzed in ImageJ.

Extracellular Acidification rate (ECAR) and Oxygen Consumption rate (OCR)

Seahorse XF96 Extracellular Flux Analyzer was used to detect cellular ECAR and OCR. On the first day, experimental and control cells were seeded into Seahorse XF96 cell culture microplates, and the XFe96 sensor cartridges were hydrated. At least 5 replicates were performed for the measurement of each group. A day later, microplates were incubated with a basic culture medium (containing 1 mM L-glutamine, without glucose) for 1 h before the assay was conducted to detect ECAR. Then ECAR was measured with the sequential injection of glucose, oligomycin, and 2-deoxyglucose (final concentration: 10 mM, 1 µM, and 50 mM respectively). For OCR detection, microplates were incubated with a basic culture medium (17 mM glucose, 1 mM sodium pyruvate, 2 mM L-glutamine, pH: 7.4) for 1 h before the assay. OCR was then measured with the sequential injection of Oligomycin, FCCP, and Rotenone/Antimycin (final concentration: 1, 1, and 0.5 µM, respectively).

Lactate Production Assay

In fresh RPMI 1640 culture media, 1 × 106 pancreatic cancer cells (SW1990 and PDX1) were co-cultured with hCAFs (human tumor-associated fibroblasts) at a 1:1 ratio for 4 h. The concentrations of lactic acid in the supernatant were measured by a lactate assay kit.

Subcutaneous Mouse Model

The Ethics Committee of Tianjin Medical University Cancer Institute and Hospital approved all proposed animal experiments, which were conducted in accordance with NIH guidelines. The pancreatic cancer cells were utilized to establish subcutaneous xenograft in 4-6-week-old C57BL/6J mice or BALB/c-nude mice, and the mice were then randomly assigned to 6 groups. Primary murine-derived fibroblasts for MME+ and MME− subpopulations were sorted, mixed with murine-derived KPC cell lines, and then subcutaneously injected into immunodeficient mice. For the subcutaneous tumor model, the indicated pancreatic cancer cells at a dilution range of 1 × 106 mixed with an equal amount of CAFs were suspended in 40 µl of PBS and then subcutaneously transplanted into each mouse’s flank. Related drugs were intraperitoneally injected one week later. The drug of interest was IOX2, with an IC50 of 22 nM. For the in vivo administration of IOX2 to mice, a dosage of 10 mg/kg was given via intraperitoneal injection once a week for a total of 3 weeks. For the AG regimen, 1000 mg/kg/week of gemcitabine (GEM) plus 300 mg/kg/week of Abraxane (AB) were injected intraperitoneally and administered once a week for 3 weeks. Immunotherapy involved the use of anti-PD1 (programmed cell death protein 1) antibodies, specifically the in vivo RMP1-14 antibody. The concentration of the solution was 2 mg/ml. Each mouse was intraperitoneally administered 200 µg once a week for a total of 3 weeks. The control group of IOX2 and AG were injected with the same volume of DMSO. The controls of anti-PD1 were injected with isotype IgG (Go in vivo RTK2758). The observers and recorders in the study were blinded to the grouping. Tumor growth was monitored every three days using a caliper, and tumor volumes were calculated by the following formula: Volume = 1/2 L1 × (L2)2, where L1 is the length of the long axis and L2 is the length of the short axis.

Functional Analysis

Gene ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis were performed to determine the biological functions of the DEGs using the cluster Profiler package. GO enrichment analysis included cellular component (CC), molecular function (MF), and biological process (BP) [33]. The parameters were configured with a significance level (p-value) of less than 0.01, a minimum count threshold of 3, an enrichment factor greater than 1.5, and a statistically significant threshold of p-value less than 0.05. The Gene Set Enrichment Analysis (GSEA) was used to enrich differentially expressed mRNA pathways, and 10,000 permutations were performed for each analysis. The KEGG Pathways dataset was selected from the curated Gene Sets. The threshold for the statistically significant GSEA analysis was set to the corrected p < 0.05 and false discovery rate (FDR) < 0.25. The reference gene set chosen was “c2.cp.kegg.v7.0.symbols.gmt”. A significance level of P < 0.05 and a false discovery rate (FDR) of less than 0.25 were used to determine significant enrichment. The result of enrichment analysis would be characterized by corrected p-values and NES. GSEA enrichment analysis and visualization were performed in GSEA local software.

Statistical Analysis

Statistical analyses were performed in GraphPad Prism-6.0. The data were reported as the mean ± standard deviation (SD) unless specified otherwise. The variance between different groups was statistically compared, and p < 0.05 was considered statistically significant. Power analysis was conducted on the results. Student’s t-test was used to compare the mean values. One-way ANOVA was carried out for the analysis of mouse tumor growth. The differential gene analysis was conducted using the top Table and decide methods. The limma package provided test functions for summarizing linear model results, performing hypothesis tests, and adjusting p-values for multiple tests. The log-rank algorithm was performed to compare statistical differences in survival curves. The Pearson test was also used to conduct correlation analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 and n.s., non-significant.